Comment by zhobbs

Possible I'm misunderstanding what you're trying to do, but ollama works well for me for local inference with qwen on my Macbook Pro (32GB).